High Spectral Spatial Resolution Synthetic HyperSpectral Dataset form multi-source fusion

TL;DR

This paper presents a synthetic hyperspectral dataset created by fusing RGB, visible, and infrared hyperspectral data to overcome individual modality limitations, enabling detailed spectral-spatial analysis for diverse applications.

Contribution

The paper introduces a novel multi-source fusion method to generate a synthetic hyperspectral dataset combining different camera modalities for enhanced spectral and spatial resolution.

Findings

Successful integration of multi-modal data into a high-quality synthetic hyperspectral dataset

Demonstrated improved spectral-spatial relationship analysis capabilities

Facilitated broader application potential in monitoring and decision-making

Abstract

This research paper introduces a synthetic hyperspectral dataset that combines high spectral and spatial resolution imaging to achieve a comprehensive, accurate, and detailed representation of observed scenes or objects. Obtaining such desirable qualities is challenging when relying on a single camera. The proposed dataset addresses this limitation by leveraging three modalities: RGB, push-broom visible hyperspectral camera, and snapshot infrared hyperspectral camera, each offering distinct spatial and spectral resolutions. Different camera systems exhibit varying photometric properties, resulting in a trade-off between spatial and spectral resolution. RGB cameras typically offer high spatial resolution but limited spectral resolution, while hyperspectral cameras possess high spectral resolution at the expense of spatial resolution. Moreover, hyperspectral cameras themselves employ different capturing techniques and spectral ranges, further complicating the acquisition of comprehensive data. By integrating the photometric properties of these modalities, a single synthetic hyperspectral image can be generated, facilitating the exploration of broader spectral-spatial relationships for improved analysis, monitoring, and decision-making across various fields. This paper emphasizes the importance of multi-modal fusion in producing a high-quality synthetic hyperspectral dataset with consistent spectral intervals between bands.